| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1961.1 | Min size FDDI packet = 17 bytes.... | NETCAD::BATTERSBY | | Tue Feb 20 1996 14:21 | 5 |
| Min size FDDI packet is 17 bytes which includes the frame control
(1 byte), 48 bit destination & source addresses (12 bytes), and
the frame check sequence (4 bytes).
Bob
|
| 1961.2 | | STRWRS::KOCH_P | It never hurts to ask... | Tue Feb 20 1996 14:37 | 5 |
|
That's interesting. That's smaller than the smallest sized Ethernet
packet which is 64 bytes, if memory serves.
|
| 1961.3 | RE: Min size FDDI packet.... | NETCAD::BATTERSBY | | Tue Feb 20 1996 16:27 | 7 |
| That's true, of course a 17 byte FDDI packet doesn't contain any
meaningful information (no data field info), yet satifies the min
requirements for an FDDI packet.
SMT packets can be smaller, but they don't get forwarded to the
Ethernet side.
Bob
|
| 1961.4 | | NETCAD::STEFANI | I *CAN* drive 65! | Tue Feb 20 1996 16:51 | 18 |
| >>That's interesting. That's smaller than the smallest sized Ethernet
>>packet which is 64 bytes, if memory serves.
I'm fairly certain that the "64 bytes" in Ethernet has to do with
maintaining signal/data integrity for a max-sized Ethernet segment where
the furthest nodes want to communicate. Since FDDI uses a ring
topology where every station retransmits the packet, the frame is "as
good" two hundred stations away as it is two stations away, thus,
FDDI doesn't have the 64 byte min packet requirement.
The smallest useful FDDI LLC packet that I've seen is a Type I 802.2
frame which is FC-DA-SA-DSAP-SSAP-Ctl-CRC which is 1+6+6+1+1+1+4 or
20 bytes in length. Bob's correct, however, that you *can* technically
design a protocol in which the mere presence of an incoming packet from
a particular source address is enough information to trigger an event.
In that case, you don't need anything beyond the MAC header.
/l
|
| 1961.5 | 20 byte LLC2 packets are useful | MARVIN::RIGBY | No such thing as an alpha beta | Tue Feb 20 1996 17:53 | 23 |
| Ethernet needed 64 bytes (actually 560us including pre-amble) of transmission to
guarantee that the most remote station had seen the packet collide. Less than
that would allow two short packets to be on the wire, travelling at considerably
less than the speed of light towards each other, and the transmitters finish
transmitting before the first bit of the other's packet arrives. FDDI to
Ethernet translation allows 802.3 frames to be sent out shorter on the FDDI than
they were on the Ethernet (no length needed on FDDI and padding removed). The
Ethernet format frames actually grow because (1) there isn't a guaranteed length
field and (2) they are 'encapsulated' in a 802.2 extended PID frame using the
Null OUI.
Doesn't FDDI also have the concept of 'short' (16-bit?) addresses, though I
don't think any digital product uses them. That would reduce the packet to
1+2+2+1+1+1+4, though it would take a couple more symbols than that on the fibre
for the frame delimiters.
I've seen FDDI packet rates quoted as 450,000pps max, that would be 56 5-bit
symbols at 125Mhz or 25ish octets per packet. (that is each-way on a Full-duplex
circuit so just a two link bridge is nearly 1 giga-packet per second for worst
case operation!)
LLC2 packets would reasonably be 1+6+6+1+1+1+4, that's a DISC frame or a UA, or
SABME. Wouldn't normally get these at high rate though.
|
| 1961.6 | | NETCAD::STEFANI | I *CAN* drive 65! | Tue Feb 20 1996 19:22 | 12 |
| >>Doesn't FDDI also have the concept of 'short' (16-bit?) addresses, though I
>>don't think any digital product uses them. That would reduce the packet to
I remember asking Paul Koning about this a long time ago and he
reluctantly said that 16-bit addressing was in the standard, but no one
used it. I believe he put it in the same light as source routing over
FDDI. :-)
The only OS specification that I've seen even mention the concept of
short addresses on FDDI is Microsoft's NDIS 3 specification.
/l
|
| 1961.7 | 17-byte packet + overhead = 446Kpps | NETCAD::ROLKE | Happy Dog is on WEMP | Mon Feb 26 1996 15:25 | 24 |
| .1 (17-byte packets) implies that a 12.5Mbyte/S link could send 735Kpps.
.5 says: I've seen FDDI packet rates quoted as 450,000pps max,
They are both right. On the actual fiber the 17-byte packet has more
overhead:
Octets What
------ ----
8 Preamble
1 Starting Delimiter
1 FC --\
6 DA \ Useful
6 SA / data (17 bytes)
4 FCS --/
0.5 Ending Delimiter
1.5 Frame Status
---- ------------------
28 A "minimum" frame
So 12,500,000 pps / 28 = 446,000+ pps.
Regards,
Chuck
|